AI Is Not Your New Employee It Is Your Intern

1 - > SPEAKER_00: Hey, welcome back to another episode of Nerds on Tap, 2 - > the podcast where we talk tech, business, and all the nerdy 3 - > stuff in between. 4 - > This is episode number 27. 5 - > Today's episode is sponsored by TKS Luxury Vacation Rentals, 6 - > located in Blue Ridge, Georgia, and Pensacola Beach, Florida. 7 - > You can find them at bestblueridge cabin rental.

com 8 - > and coming soon in a couple weeks, not live yet, 9 - > bestpensacola beachrental.com. 10 - > Today's episode is all about AI and not the sci-fi version. 11 - > Come on, folks.

12 - > We're talking real-world boots on the ground AI for small 13 - > businesses. 14 - > So today's episode will be titled AI for Small Business, 15 - > What's Real, What's Risky, and What's Straight Up Hype. 16 - > And joining me in person today, a very rare find in Pensacola, 17 - > all the way up from beautiful Lakeland, Florida, known for its 18 - > roaming swans. 19 - > We have Brian Wilkie, the VP and Chief Operating Officer of 20 - > Digital Boardwalk.

21 - > Welcome to the show, Brian. 22 - > Thank you. 23 - > Thanks for having me. 24 - > So, Brian, are you prepared to talk about AI?

25 - > We will see. 26 - > So we're going to start off with AI for small business, helpful 27 - > versus hype versus hard no. 28 - > Okay. 29 - > So my question to you, let's start broad.

30 - > We'll make it easy. 31 - > I'm going to take it. 32 - > We don't share the show notes, folks. 33 - > So, you know, don't be nervous.

34 - > Sure. 35 - > Uh AI is everywhere right now. 36 - > Right. 37 - > So, what are business owners getting completely wrong?

38 - > SPEAKER_01: Um not taking action. 39 - > I think uh there's a lot of businesses who are seeing this 40 - > as hype or a bubble and they're choosing to stay on the 41 - > sidelines and see where it lands. 42 - > Um I don't think there's a question at this point that AI 43 - > is going to play an important role in business moving forward. 44 - > And we don't know what that's going to look like yet.

45 - > It's still evolving, but the businesses who are sitting on 46 - > the sidelines and aren't at least playing around with it to 47 - > learn what's happening in the world. 48 - > SPEAKER_00: Kind of the fundamentals, right? 49 - > Right. 50 - > How to use it, what kind of answers you're going to get out 51 - > of it, what what it can do, what it can't do.

52 - > SPEAKER_01: Right. 53 - > Yeah, because their competition is. 54 - > And when this technology begins to mature, the ones who are 55 - > playing around with it and had that head start are going to be 56 - > able to implement it quickly. 57 - > Whereas everyone else is going to be playing catch up.

58 - > And as everyone sees the pace of AI move so quickly that if 59 - > you're playing catch up, you're going to lose. 60 - > SPEAKER_00: So what's the biggest misconception do you 61 - > think small businesses have about AI, you know, right now, 62 - > this very moment? 63 - > What as a small business owner, um, you know, when you think 64 - > about AI, when someone's actually fooling around with it, 65 - > fiddling around with it, like you say, learning about it. 66 - > Yeah.

67 - > What is their misconception? 68 - > Is it is it going to completely replace all their staff? 69 - > Highly doubtful for small businesses. 70 - > SPEAKER_01: Um it's not a magic employee.

71 - > No, it's um as far as job replacement goes, I see that 72 - > happening in the enterprise space because historically, if 73 - > uh an enterprise had a problem to solve, they would throw 74 - > people at the problem. 75 - > And they're not gonna have to do that anymore. 76 - > So I think we're gonna continue to see mass layoffs in the 77 - > enterprise space. 78 - > Small businesses are gonna use it as a an optim optimization 79 - > and and scalability.

80 - > SPEAKER_00: So it's not a magic employee, it's more like a a 81 - > fast intern. 82 - > SPEAKER_01: Yes, yeah, and that might be the other, you know, 83 - > misconception for businesses that discourages them from AI is 84 - > they see the news promoting that AI can do all of the work, it 85 - > can do 100% of the coding or what whatever they advertise, 86 - > and then they get into it and they see it doesn't do that, and 87 - > their gut reaction is okay, it's just hype, they're trying to 88 - > sell me products.

89 - > SPEAKER_00: Right. 90 - > And we're at the beginning, you know, this is just a start. 91 - > Yeah. 92 - > It it's it's going to rapidly, rapidly evolve.

93 - > Yeah, and we're gonna see that over the next two or three 94 - > years, probably sooner than later. 95 - > Um, so where are you actually seeing AI save real time for the 96 - > small business owner today, not just in theory, but real 97 - > implemented practices built around AI. 98 - > What is it? 99 - > How is it helping the the typical small business owner?

100 - > How can they utilize it outside of reaching out to Digital 101 - > Boardwalk and having Boardwalk help them? 102 - > Yeah. 103 - > SPEAKER_01: Um this is the one thing that's changing and has 104 - > changed within the last few weeks. 105 - > So my answer is not going to age very well.

106 - > But right now, as far as what small businesses are using, um, 107 - > I'm seeing it really help them in the research and planning 108 - > space. 109 - > Right. 110 - > Yeah. 111 - > So as they're making business decisions, if they're trying to 112 - > understand their market, um, trying to make some high-level 113 - > business decisions, you could have previously gone to search 114 - > engines to get some of that data.

115 - > It would just take you forever to do it. 116 - > But it's fragmented. 117 - > SPEAKER_00: Yes. 118 - > When you go out on Google, right?

119 - > Right. 120 - > But AI is searching so much more and it's bringing all that data 121 - > in, and then it's learning from it if you m stay in the same 122 - > chat, right? 123 - > Mm-hmm. 124 - > SPEAKER_01: Yeah, or or the same platform, right?

125 - > So whatever you use, Microsoft Copilot, Gemini, yeah, Claude. 126 - > SPEAKER_00: Yeah. 127 - > So that's interesting. 128 - > I mean, I've seen it help me with high-level uh concepts 129 - > around business that I research and development.

130 - > Yeah. 131 - > I mean, if if somebody sends me an NDA prior to punting it off 132 - > to my attorney, I might do a preliminary on it using AI, and 133 - > it'll give me some suggestions. 134 - > And then, you know, if if everything looks good to go, I 135 - > don't have to involve my you know,$400 an hour,$500 an hour 136 - > attorney. 137 - > So that leads me to the next thing, and that's just one 138 - > thing.

139 - > I mean, calendaring and and being able to collaborate 140 - > between emails and and and the whole Microsoft ecosystem, which 141 - > I've seen, it's been a game changer for me. 142 - > But, and this isn't one of my questions, but talking about 143 - > NDAs, I mean, law firms, you know, are they worried? 144 - > Should they be worried? 145 - > I mean, you're you're seeing, I mean, not just law firms, but 146 - > any any sort of outside.

147 - > So, how is that gonna impact it, you think? 148 - > SPEAKER_01: Yeah, so uh you you'll see this across other 149 - > conversations with experts in the space, but AI is really 150 - > gonna hurt uh cognitive repetitive work. 151 - > Um, eventually it's also gonna impact manual repetitive um with 152 - > robotics. 153 - > But yeah, uh contract review analysis, a lot of the junior 154 - > level legal work.

155 - > So if we're talking law firms, junior level legal work um is 156 - > going to be displaced with AI. 157 - > So the concern there is all the kids going to school for legal 158 - > right now, coming out of college, they're not gonna have 159 - > many job opportunities. 160 - > Got it. 161 - > SPEAKER_00: So we still need our attorneys.

162 - > We're still gonna need senior levels. 163 - > I like my attorney. 164 - > Yeah. 165 - > So, you know, that leads me into my next question because I mean, 166 - > let's stay on this topic.

167 - > Um talking about low-level work, talking about some other things. 168 - > What's something business owners are trying to use AI for that 169 - > makes you immediately go, that's a bad idea? 170 - > Hmm. 171 - > SPEAKER_01: Uh well, um, so a great example of this, um, I was 172 - > actually at a a Microsoft event recently, and there was a 173 - > question in the audience, and they said, We know Microsoft 174 - > Copilot runs the GPT models from OpenAI and now the anthropic 175 - > Claude models.

176 - > Uh, but when they prompt Chat GPT or Claude outside of the 177 - > Microsoft platform, they get different responses than Copilot 178 - > gives. 179 - > And the reason for that is Microsoft is very focused on the 180 - > risk side of things. 181 - > They're very risk averse with how they're putting up their 182 - > guardrails. 183 - > And so if you go to ChatGPT and feed it six months worth of 184 - > performance appraisals for an employee and ask it, should we 185 - > terminate this employee based on the performance appraisals?

186 - > ChatGPT is going to be like, yeah, you probably should. 187 - > Wow. 188 - > Whereas Copilot is going to say, eh, here's some insights based 189 - > on these performance appraisals, but I can't make that decision 190 - > for you. 191 - > Microsoft has specifically coded the platform to behave that way 192 - > because of the legal implications of leaning on 193 - > something like AI to make those types of decisions.

194 - > Um people doing those types of things, um, or just feeding 195 - > sensitive data into platforms. 196 - > You know, we're gonna hear Claud's the greatest thing, and 197 - > then feeding your whole customer list into it. 198 - > SPEAKER_00: You're gonna shorten this episode if you don't let me 199 - > get to that topic. 200 - > Security is a big one.

201 - > Yeah, and we're gonna end, I think we're gonna we're gonna 202 - > get to that in a little bit because that is a big hot topic. 203 - > Obviously, we both work at digital boardwalks, so uh we 204 - > know all about security and and I want to get into that. 205 - > But you know, so if you had to draw a line, for instance, we're 206 - > talking about you know, drafting something versus actually using 207 - > it. 208 - > So if we're talking about legal, and you're you know, I talked 209 - > about NDA.

210 - > If it's just something small and I just need a little touch 211 - > point, yeah, yeah, I might run with it because I know enough 212 - > about how that works, and it's an NDA. 213 - > Yeah. 214 - > I mean, you want to cover you know confidentiality and and and 215 - > and various things in there, yeah um, but certain things 216 - > might not matter depending on who you're engaging with on the 217 - > NDA. 218 - > But if it's super deep, you want to punt that.

219 - > You know, you want to punt that to legal. 220 - > So, where do you draw the line between drafting help and risky 221 - > shortcut? 222 - > I mean, a risky shortcut would be like, eh, I'm just gonna do 223 - > it myself, punt it over. 224 - > Yeah.

225 - > SPEAKER_01: Um, I I think you kind of set this up perfectly 226 - > the way you explained it before. 227 - > Look at it as an assistant. 228 - > So maybe not a subject matter expert per se, but someone who 229 - > can help get you started, lay the groundwork, get get it 80% 230 - > of the way. 231 - > Yeah.

232 - > Um, and don't just explicitly trust it. 233 - > SPEAKER_00: Um, so so let me end this segment by asking one final 234 - > question. 235 - > I mean, we get approached about AI at digital boardwalk all the 236 - > time, and we're helping uh companies now, and obviously 237 - > we're still in the early phases of all this, but it it's going 238 - > to evolve. 239 - > If a 10-person company asked you, and you already answered 240 - > alluded to this earlier, but if they asked you, should we be 241 - > using AI, what's your honest answer and where would you point 242 - > them to begin?

243 - > SPEAKER_01: Yes, definitely start. 244 - > Um, and start with the whole team. 245 - > Uh, we we see a lot of companies try to baby step this and say, 246 - > all right, we're just gonna have our executive team play with the 247 - > tools. 248 - > No, let give the tools to the whole team so everyone can start 249 - > getting that experience.

250 - > Um, but start simple with chat. 251 - > That that is the easiest way to get started. 252 - > You there's very little learning curve to it. 253 - > Um the way we did it internally, back when this stuff first 254 - > started launching, we just made the chat everyone's default 255 - > homepage in the browser.

256 - > So instead of them going to Google, they would just prompt 257 - > AI for the the question they had. 258 - > Um start there when people start to see how that can help save 259 - > time and gather information quicker. 260 - > They start to explore naturally and then they ask questions 261 - > like, okay, what else can this do? 262 - > And it's that curiosity that's gonna really help businesses as 263 - > this evolves.

264 - > Good answer. 265 - > SPEAKER_00: All right, let's let's use that as a segue in the 266 - > segment two. 267 - > We're gonna talk about AI at work, automating the boring 268 - > stuff. 269 - > Yeah.

270 - > Um, you know, and that's that's part of what you do here at 271 - > Digital Boardwalks. 272 - > So this is a good segue. 273 - > So we'll kick it off with the first question. 274 - > Where do you look when you look at a typical work day?

275 - > What are the boring tasks AI is actually really good at taking 276 - > off someone's plate? 277 - > SPEAKER_01: Anything repetitive. 278 - > If you do something the same way every time with any sort of 279 - > frequency, and you can clearly communicate that or or write 280 - > that process, AI is gonna be fantastic for that because you 281 - > can define very clear actions and boundaries and it can it can 282 - > run with it. 283 - > SPEAKER_00: So a first draft of maybe an email, but not 284 - > necessarily the final draft.

285 - > Sure. 286 - > Yep. 287 - > Because you gotta put you gotta humanize it. 288 - > SPEAKER_01: Yeah.

289 - > SPEAKER_00: And and I can always tell when I get an email that's 290 - > written by AI. 291 - > SPEAKER_01: Yeah. 292 - > So, and this starts getting into some of the nuance of it that 293 - > people who aren't into tech really don't uh discover without 294 - > guidance, but you can steer AI to respond in ways that you want 295 - > it to. 296 - > Um, so the telltale sign that something's AI generated or 297 - > those long dashes, um, that's just the dead giveaway because 298 - > they all use it.

299 - > Um, you can customize your settings so it explicitly 300 - > ignores that behavior or it doesn't output that way. 301 - > And instead, you can say, hey, give me bullet points or use 302 - > more complex sentence structure, those types of things. 303 - > So you can make it feel more natural, but you have to know 304 - > how to give it that guidance. 305 - > SPEAKER_00: So, in the world of cybersecurity, and this isn't 306 - > about cybersecurity, folks, but since I've got you in here, is 307 - > that what the bad actors are doing with their their uh their 308 - > phishing emails?

309 - > Absolutely. 310 - > Trying to humanize them by by using those tricks. 311 - > SPEAKER_01: Oh, yeah, for sure. 312 - > For sure.

313 - > Uh, you know, what 10 years ago, phishing emails, people were 314 - > always coached, like, look for the spelling errors, look for 315 - > the weird language and the sentence structure and that kind 316 - > of stuff. 317 - > That's gone now. 318 - > AI can perfectly draft those emails to be very legitimate 319 - > looking. 320 - > SPEAKER_00: Um, and it it does that very well.

321 - > So, what what's what's a workflow that you can think of 322 - > that you would never automate end to end? 323 - > Oh I know, right? 324 - > I um I threw this one in there last minute because I wanted to 325 - > see if I could get you. 326 - > SPEAKER_01: The most risky one I can think of, and and this is a 327 - > little ambiguous, but I would say mass actions.

328 - > Those are dangerous. 329 - > Okay. 330 - > Um if you well, I mean, if you don't put enough thought into 331 - > it, you could destroy a system, you could destroy the data that 332 - > you've built, you know, spent years building. 333 - > You know, if you've got a customer database and you can 334 - > say right it's gonna write over everything.

335 - > Right, yeah. 336 - > Or the I forget what organization it was. 337 - > There was a a fairly large enterprise recently that uh 338 - > their their coders leaned on AI to make a change in the code 339 - > base, and it wiped out the entire code base. 340 - > So it did more harm than good, and then they had to all go back 341 - > and try to to fix what it had done.

342 - > Wow. 343 - > Um, so yeah, that's a that's a buzzkill. 344 - > Yeah. 345 - > So yeah, I think the danger is in any sort of mass actions, um, 346 - > at least without some sort of checks and balances or 347 - > verification before it performs the actions.

348 - > SPEAKER_00: Right. 349 - > So it's not a hundred percent replacing humans. 350 - > SPEAKER_01: This is where I think the language gets 351 - > misconstrued because you'll see big platforms, anthropic 352 - > included, saying it's doing 90% of our code or it's doing 100% 353 - > of our code. 354 - > That doesn't mean it's uh doing 100% of the work for the 355 - > employee, it's just doing uh that legwork first.

356 - > So is the code messy? 357 - > It can be, it can also be clean. 358 - > It it depends on your coders. 359 - > Um, some coders are really messy coders, so it could be better 360 - > than some of them.

361 - > But what I think what they mean specifically with code isn't 362 - > there a movie where they uh uh it's it's uh spaces versus 363 - > Silicon Valley. 364 - > SPEAKER_00: Oh, tabs versus spaces, yeah. 365 - > SPEAKER_01: I'm a tab guy that's funny, but uh I I mean, even for 366 - > myself, you know, three, four years ago when I'm coding, I'd 367 - > be writing it all manually. 368 - > Yeah.

369 - > Now it's rare that I write any code directly. 370 - > I'll use AI to help generate it. 371 - > I'll go in and modify it, but it's it's rare that I'm just 372 - > sitting there at the command line. 373 - > SPEAKER_00: I've increased my productivity by at least at 374 - > least 25% by using AI, maybe maybe even more than that.

375 - > Yeah, because I'm not spending as much time spinning wheels, 376 - > trying to either find the resources, look for what I need, 377 - > or make it do what I need it to do. 378 - > So it's really so that brings me to my next question. 379 - > What's a real example? 380 - > Small business owners pay attention to this question.

381 - > What's a real example you've seen where AI made an employee 382 - > noticeably, noticeably more productive? 383 - > SPEAKER_01: This is I'll speak from first hand experience. 384 - > Are we going to talk about edge? 385 - > Did we make it more productive with AI?

386 - > No, I'll I'll share first hand experience because this was an 387 - > aha moment for me from an AI perspective. 388 - > Um up until two or three weeks ago, AI has been really good at 389 - > giving me answers. 390 - > It was not very good at doing work. 391 - > It's the one thing that always disappointed me with it.

392 - > There was so much hype around AI, and I'd sit in front of 393 - > something like Excel and say, Hey, do this for me. 394 - > And it couldn't do it. 395 - > It could it could help me. 396 - > It could give me guidance on how to build a formula or how to set 397 - > up the pivot table, but it it wouldn't do the work.

398 - > That's changed. 399 - > Um, AI is is going agenc, as you'll hear, uh agent-based. 400 - > And all that really means for the layman is it can do the work 401 - > now, it can do the tasks. 402 - > And so when Excel became agentic a few weeks ago, let's step back 403 - > for our audience.

404 - > SPEAKER_00: Yeah. 405 - > Agent-based. 406 - > It can do the task, right? 407 - > What does that mean, agent-based for our audience?

408 - > SPEAKER_01: Yeah, so it is it's a means of connecting apps and 409 - > services to uh uh permissions that give it the ability to to 410 - > perform tasks in those apps. 411 - > Okay. 412 - > So pre prior to this, it was surface level. 413 - > It could see information, but it couldn't do anything with it.

414 - > It could maybe analyze it, it could give you guidance, but it 415 - > couldn't perform any actions within the app. 416 - > Um now it can. 417 - > And so yeah, when when the agent landed in Excel, I was working 418 - > on uh a financial model, and there was a column in my in my 419 - > worksheet that was functional, but I could tell based on the 420 - > results that there was a variable that wasn't taking into 421 - > account in the formula. 422 - > I needed it to also include this other condition.

423 - > And of course, I could have figured that out. 424 - > I could have looked at the formula and made that 425 - > modification. 426 - > But at the same time, as this financial model and this data 427 - > set was starting to grow, I wanted to move everything from 428 - > this worksheet to a different worksheet. 429 - > Um and for any Excel nerds out there, you know you need to like 430 - > anchor your references with the dollar sign if you want to do 431 - > moves like that, which I hadn't.

432 - > I hadn't planned ahead for this one. 433 - > And then after all of that, I also wanted to format it a 434 - > certain way just to make the data a little more readable. 435 - > So I saw agent mode, turned it on, and I was like, let me just 436 - > give this a shot. 437 - > So I gave it the prompt.

438 - > I said, I need this column to also take into account. 439 - > This data variable. 440 - > I want you to move it to a different worksheet, name it 441 - > this, and then I want you to reformat the data and then 442 - > create a pie chart based on that data. 443 - > And I hit go and sat back and watched it.

444 - > And what's really cool, when you when you chat with AI, you you 445 - > wait for a moment and then it just kind of spits everything 446 - > out. 447 - > It does most of the work behind the scenes. 448 - > In agent mode, you watch it step by step. 449 - > So I saw it break my formula first, then I saw it add in the 450 - > new component.

451 - > It brought in the results. 452 - > I saw it create the new tab, name it, move my columns over, 453 - > build out the formatting, add the chart. 454 - > It even caught itself and said, okay, the chart has some data 455 - > label issues. 456 - > So let me fix that.

457 - > Interesting. 458 - > And so something that I probably could have knocked out in about 459 - > an hour, it took maybe like two minutes. 460 - > SPEAKER_00: So the first time I used Claude, I had it design a 461 - > simple website for me. 462 - > I watched the code.

463 - > I watched it code right in front of. 464 - > Yep. 465 - > Is that what you're referring to? 466 - > SPEAKER_01: Exactly.

467 - > And in my experience with Excel and Microsoft Copilot, it's 468 - > powered by Anthropics Claude. 469 - > SPEAKER_00: Anthropics Claude. 470 - > Yep. 471 - > Got it.

472 - > So we're talking about being more productive with AI. 473 - > Let's flip that upside down. 474 - > Let's put it, let's put them on its head. 475 - > What are people trying to automate with AI?

476 - > That it just completely fumbles. 477 - > SPEAKER_01: Okay. 478 - > So this gets into the world of process automation. 479 - > And you can't automate a process that you don't have well defined 480 - > already.

481 - > You need to understand how that process should work so that when 482 - > you are automating it, you can make sure that you're getting 483 - > the output you expect. 484 - > Um, because otherwise you're just trusting what it's doing, 485 - > and you have no idea what that output means. 486 - > SPEAKER_00: Most of the audience probably isn't going to be 487 - > automating with it yet. 488 - > No.

489 - > They're going to be using things like writing emails and things 490 - > like that. 491 - > So context and consequences matter. 492 - > So let's say, let's use a uh real world scenario where maybe 493 - > you have a frustrated customer and you're trying to respond to 494 - > that frustrated customer using AI, uh, you know, requires 495 - > accountability. 496 - > Yep.

497 - > How is AI going to address that frustrated customer? 498 - > And and can you tell it, hey, be direct? 499 - > Hey, we had had to do that one day, or be a little bit more uh 500 - > passionate about their plight, or you know, or whatever they're 501 - > frustrated about. 502 - > Can we do that now?

503 - > Could we do that? 504 - > Could we do that a year ago? 505 - > What how has it evolved? 506 - > And in and you know, let's let's just talk about that.

507 - > SPEAKER_01: Yeah. 508 - > Um, I think the biggest thing that's happened over the last 509 - > year. 510 - > So, in short, yes, you can do that. 511 - > Um, what's happened over the last year is the the information 512 - > that AI has access to to generate its responses is 513 - > expanding.

514 - > We're able to tie more things in together. 515 - > And so that has always been the crux of it in the early days is 516 - > it didn't have enough context to give you a valuable response. 517 - > It it was just a generalist. 518 - > It would be based on whatever information is out there on the 519 - > web.

520 - > Um now that it's able to tie into your emails, the way you 521 - > treat other customers, the way you talk to people internally, 522 - > um, it can use that tone and and way that you approach conflict. 523 - > So if you're mean to customers, AI is gonna address us. 524 - > SPEAKER_00: Are you kidding me? 525 - > Yeah, it will follow lead.

526 - > Wow. 527 - > Yeah. 528 - > Don't be mean to customers, folks. 529 - > Be nice.

530 - > Yeah. 531 - > Uh so we're not replacing roles, we're helping people be more 532 - > productive. 533 - > Uh, if someone on your team uses AI every day, right? 534 - > Uh, what's the difference between them doing it the right 535 - > way and the wrong way?

536 - > Oh, and we'll wrap up this segment with that question. 537 - > SPEAKER_01: Yeah, that that easily just comes down to 538 - > trusting the output. 539 - > Right. 540 - > Like just putting what we just talked about.

541 - > Right. 542 - > Putting in a prompt, you take what it gives back and you run 543 - > with it. 544 - > That is not a good idea. 545 - > Um even just having a a couple quick follow-up prompts to tune 546 - > it, to clarify it, can make a world of difference.

547 - > SPEAKER_00: Well, you you alluded to security earlier. 548 - > I told you to hang on, don't get ahead of us. 549 - > We're gonna jump into that segment now. 550 - > So the next segment is the AI security checklist.

551 - > And I know I have a weird feeling this could take an hour 552 - > because we are very passionate about security and and 553 - > understanding the concepts of security and helping our 554 - > customers address those security concerns. 555 - > Yeah. 556 - > So my first question to you is before we even talk about AI, 557 - > before we even get into that, what are the basic security 558 - > things every small business needs to still need to have 559 - > locked in? 560 - > AI aside, the stuff we already talked to our customers about, 561 - > what security checklist?

562 - > What do they need to have locked in? 563 - > Always. 564 - > SPEAKER_01: Yeah. 565 - > Before you can even really go down the AI journey, you have 566 - > got to These are the non-negotiables.

567 - > Yeah. 568 - > You you've got to focus on your data governance. 569 - > You need to have control over who has access to what. 570 - > Right.

571 - > Um permissions. 572 - > Yep. 573 - > SPEAKER_00: Permissions. 574 - > MFA.

575 - > Is that should we be selective? 576 - > Non-negotiable. 577 - > Is digital boardwalk selective on who gets MFA? 578 - > SPEAKER_01: No, that is mandatory.

579 - > SPEAKER_00: Okay. 580 - > SPEAKER_01: Um, yeah, and that one, you know, sadly, it's you 581 - > know, there's ways around that too. 582 - > Yeah. 583 - > Um, so for us, um, we're also going full zero trust.

584 - > Um, I would really consider that a non-negotiable moving forward. 585 - > What is zero trust, Brian? 586 - > Um, so try not to go too long-winded to you. 587 - > SPEAKER_00: So that's if if if if I don't trust my kids to do 588 - > anything, I'm not going to allow them to do anything.

589 - > Is that what that means? 590 - > Sort of. 591 - > SPEAKER_01: So historically in cybersecurity, we would define 592 - > what bad looks like. 593 - > Let the systems try to identify when bad happens and then try to 594 - > stop it.

595 - > Uh, zero trust is the reverse of that. 596 - > So we define what good looks like, and then anything else is 597 - > blocked. 598 - > SPEAKER_00: So I plug a flash drive into my laptop right now 599 - > that is an untrusted device. 600 - > Yep.

601 - > My laptop's a trusted device, but the flash drive isn't. 602 - > Yep. 603 - > Flash drive's blocked. 604 - > Is oh wow.

605 - > Yep. 606 - > So I can I can't access uh my picture to the beach. 607 - > Nope. 608 - > Okay.

609 - > SPEAKER_01: And take it a step further. 610 - > Let's say we did previously trust that device and you go 611 - > home and unintentionally get a piece of malware on that uh USB 612 - > device and you plug it back into your work computer and try to 613 - > open it. 614 - > We don't trust that piece of malware that's on the drive. 615 - > SPEAKER_00: So that gets blocked too.

616 - > Yeah, that's amazing. 617 - > Yep. 618 - > So those are the those are the that's the way you want to take 619 - > it. 620 - > What's the biggest security mistake customers are making, or 621 - > not customers, but our customers don't make them because they 622 - > hire us.

623 - > But what's the biggest security mistake companies are making 624 - > right now when it comes to using AI tools? 625 - > Uh we talked about this earlier. 626 - > You talked about not plugging in sensitive information to public 627 - > chat. 628 - > Yeah.

629 - > Right? 630 - > Yeah. 631 - > SPEAKER_01: I yeah, I I think it's not having a an AI security 632 - > conversation in the business. 633 - > They're just jumping in, letting everyone play with whatever 634 - > tools they want, and your your non-technical people aren't 635 - > thinking about what it means for them to put sensitive 636 - > information.

637 - > SPEAKER_00: So I can't take our entire company's payroll, plug 638 - > it in the chat GPT, and ask it if if if that's you know, if 639 - > anything should be changed. 640 - > SPEAKER_01: No. 641 - > SPEAKER_00: Don't what happens to that when it goes out into 642 - > ChatGPT land? 643 - > SPEAKER_01: Yeah, that data is used to train the model and can 644 - > also be used in responses to other companies.

645 - > SPEAKER_00: So a competitor could use that data against you. 646 - > Yep. 647 - > Okay. 648 - > Um okay.

649 - > So wow. 650 - > So no financials, definitely don't plug passwords in. 651 - > No. 652 - > Right?

653 - > Yeah. 654 - > SPEAKER_01: And be careful of what AI apps you're downloading. 655 - > SPEAKER_00: We're seeing this a lot too, where um personal 656 - > identifiable uh fiable information, none of that, 657 - > right? 658 - > PII.

659 - > Yep. 660 - > Um go ahead. 661 - > Sorry. 662 - > SPEAKER_01: Yeah, so bad actors are are riding this wave of AI 663 - > hype.

664 - > So they're putting out a bunch of fake AI agents that people 665 - > are using to feed in sensitive information like that, and 666 - > they're just using using it to collect it and then sell it on 667 - > the dark web. 668 - > Sell it on the dark web. 669 - > SPEAKER_00: Oh my gosh. 670 - > Yep.

671 - > So let's say let's say a business wants to incorporate AI 672 - > safely. 673 - > Okay. 674 - > What's your simple non-negotiable checklist? 675 - > If if they want to put it in safely, what what shouldn't they 676 - > do?

677 - > SPEAKER_01: Well, don't let your employees have the freedom to 678 - > use whatever AI platform they want. 679 - > Right. 680 - > You need to clearly define that. 681 - > Um, and just have an honest conversation with the team about 682 - > the risks of AI, it being in early development.

683 - > We need to think about how we're using these these tools and how 684 - > we're protecting our company information. 685 - > Got it. 686 - > SPEAKER_00: So how how big of a problem, and and you can define 687 - > this term shadow AI, but how how big of a problem is shadow AI? 688 - > Uh, employees using tools without approval, and how do you 689 - > even control that?

690 - > Like how do you manage it? 691 - > SPEAKER_01: Yep. 692 - > So it's a massive problem right now. 693 - > Um, and zero trust is the approach for it.

694 - > Okay. 695 - > SPEAKER_00: Um, so but that's only on the machines you 696 - > control, right? 697 - > Correct. 698 - > Yeah.

699 - > So they could still go on a personal machine and yeah. 700 - > How do you can is there a way to control that yet? 701 - > SPEAKER_01: Yeah. 702 - > So you you extend zero trust to the point where you you you put 703 - > up a barrier around the information itself.

704 - > Okay. 705 - > So if let's say all of your data lives in Microsoft 365, you 706 - > prevent an employee from accessing their Microsoft 707 - > account, their 365 account, from any personal devices that aren't 708 - > trusted by the organization. 709 - > So there's no way for them to extract the business data out to 710 - > put it into whatever random AI chatbot they downloaded. 711 - > SPEAKER_00: Incredible.

712 - > So let's finish on one last question in that segment. 713 - > We'll jump to the final segment. 714 - > Um, what's one thing you would absolutely tell a business? 715 - > Do not put this into AI under any circumstance.

716 - > I think we already alluded to that. 717 - > Passwords. 718 - > SPEAKER_01: Anything you don't want the rest of the world to 719 - > see. 720 - > Yeah.

721 - > That's the easiest way to think about it. 722 - > SPEAKER_00: Yeah. 723 - > Um, all right. 724 - > So we're gonna use that as a segue to segment for.

725 - > Um, we'll call this your employees are already using AI, 726 - > right? 727 - > Um, so I mean, let's be honest, they're using AI whether you 728 - > like it or not. 729 - > Yeah. 730 - > So if you haven't adopted it yet, I guarantee you at least 731 - > one of your employees has already adopted AI and is using 732 - > it.

733 - > So, how many employees are already using AI tools, would 734 - > you say if you had to put a number on it, um without the 735 - > company even realizing it? 736 - > All of them. 737 - > SPEAKER_01: Um so some of the larger companies, Google, uh, 738 - > Microsoft, they like to boast this because they kind of 739 - > shoehorned in AI into their search engines. 740 - > So now every time you search something, that top result is AI 741 - > generated.

742 - > Yeah. 743 - > So then they can make the claim that we've got how how many 744 - > millions of AI users. 745 - > But the world knows about it at this point. 746 - > You know, all your employees know about it.

747 - > And if you ask them, I guarantee most of them have probably 748 - > downloaded and used Chad GPT. 749 - > Whenever that you know hit the landscape, everyone was curious 750 - > about it. 751 - > So um, they're all using it. 752 - > Um how they're using it is the concern.

753 - > Think kids are using it for homework? 754 - > For sure. 755 - > Yeah. 756 - > Would you have?

757 - > Probably. 758 - > Okay. 759 - > Probably. 760 - > At least uh, you know, help me in the process.

761 - > There you go. 762 - > SPEAKER_00: Yeah. 763 - > Um, all right. 764 - > So what are the biggest risks?

765 - > We we just talked about this, but you know, using chat GPT or 766 - > Copilot on their own, what are the biggest risks when employees 767 - > start using those tools with m without maybe uh mandated 768 - > company policies built around that? 769 - > So what where are the risks? 770 - > I mean, if if if if if John Doe in in Department A, you know, is 771 - > using Chat GPT while he's at work, not not managed, and he's 772 - > plugging stuff in, you know, maybe maybe there's no ill 773 - > intent, maybe he's doing everything, you know, but but 774 - > what are the what's the downside to that?

775 - > I mean, the true downside to him doing that, even though he 776 - > hasn't done anything wrong, yeah, he hasn't plugged his 777 - > passwords or financials or payroll in, like we talked 778 - > about. 779 - > And and he has no ill intent. 780 - > He's just trying to do a good job during the day, and but he's 781 - > using it on the side to assist him. 782 - > There's no mandated company policies in place for that.

783 - > Yeah. 784 - > But how do we how do we get control of that? 785 - > SPEAKER_01: Uh it again, it comes back to conversations as a 786 - > company and then implementing technical controls to to limit 787 - > it. 788 - > You know, the the way you frame that, there probably isn't an 789 - > issue with it in the beginning.

790 - > Um, but there's nothing stopping them from taking it that one 791 - > extra step, which crosses the line. 792 - > Um and so putting controls in place before that line is 793 - > crossed is really important. 794 - > What about banning it? 795 - > Don't want to ban it.

796 - > Why not? 797 - > It it just forces people to use it anyway. 798 - > SPEAKER_00: Yeah. 799 - > Um, it's just like kind of like telling your kids no.

800 - > SPEAKER_01: Exactly. 801 - > SPEAKER_00: It makes them want to do it more, right? 802 - > Right? 803 - > Yeah.

804 - > So uh so it's not realistic to me. 805 - > No, no, and I would never encourage, I mean, we're we're 806 - > IT, right? 807 - > We're going, we're early adopters of it, right? 808 - > So banning it's just gonna turn into shadow IT or or shadow AI, 809 - > I'm sorry.

810 - > Uh safer, safer ways, approved tools, plus simple rules, 811 - > training, accountability. 812 - > Um what's the simplest policy language you've seen that 813 - > actually gets followed? 814 - > And it doesn't have to be for AI, it can be across anything, 815 - > but how would you apply that to AI? 816 - > SPEAKER_01: Yeah, um operationally for most small 817 - > businesses, it really just comes down to an approval process.

818 - > So they'll they'll define a like a chain of command to authorize 819 - > where they can use something like that, and at least that 820 - > forces the conversation. 821 - > SPEAKER_00: So look into your camera over here and tell that 822 - > small business owner what he can do right now. 823 - > If they're not, if he if he, you know, he knows some people might 824 - > be using AI, but what he could do right now, like in in terms 825 - > of guidelines, to start to get it moving inside of his 826 - > organization and have those conversations with his employee.

827 - > What is he gonna do? 828 - > SPEAKER_01: Yep. 829 - > So meet with your whole team, encourage them to use AI and 830 - > explore, but before they begin using a new tool that they want 831 - > to explore, have them approach you first, if assuming you're 832 - > the decision maker in this, and just have that one extra step, 833 - > that one conversation about is this secure? 834 - > How do we protect our information?

835 - > How could this be exploited? 836 - > Um, you don't have to necessarily involve IT yet, just 837 - > having that extra step will prevent people from just running 838 - > free and using the tools. 839 - > SPEAKER_00: So they need an AI policy. 840 - > Yep.

841 - > Um, and it's not overkill to have that. 842 - > No. 843 - > Okay. 844 - > Because people are already using it anyway.

845 - > Yeah. 846 - > All right. 847 - > So what's what's the what's the risk of not having that policy 848 - > in place? 849 - > Well, then it's it's a free-for-all.

850 - > SPEAKER_01: It's free for all. 851 - > Yeah. 852 - > And there and you don't have any way of knowing when it's 853 - > happening either. 854 - > SPEAKER_00: Oh, so let's boil that boil down that policy.

855 - > What's one thing that that policy actually absolutely has 856 - > to have in it? 857 - > Probably framework, right? 858 - > What what during Again? 859 - > SPEAKER_01: I would just base it on an approval process.

860 - > Like if you want to use a new tool that isn't authorized, you 861 - > need approval to do that first. 862 - > SPEAKER_00: But how do you build a policy that they're gonna 863 - > follow? 864 - > I mean, it's other than just making a policy and putting it 865 - > in a folder outside of actual IT controls, it's training. 866 - > Right.

867 - > Training, yeah. 868 - > Repeat, repeat, repeat. 869 - > Right? 870 - > And then remind, remind, remind.

871 - > Yeah. 872 - > Okay. 873 - > SPEAKER_01: Um has to be in the conversation. 874 - > SPEAKER_00: All right.

875 - > Well, ladies and gentlemen, we're gonna go into the rapid 876 - > fire round with Mr. 877 - > Brian Wilkie here. 878 - > Uh Brian, you've listened to the show before, you know how this 879 - > works. 880 - > Uh, we're gonna ask you one one, you know, simple questions, a 881 - > bunch of them.

882 - > Okay. 883 - > And you're gonna answer with the first word that comes to your 884 - > head. 885 - > Okay. 886 - > So if I say, who do you dislike?

887 - > Don't say Tim Shoot. 888 - > Yeah. 889 - > Um, so we're gonna go. 890 - > Ready?

891 - > Let's do it. 892 - > All right, here we go. 893 - > Uh AI you actually use every day. 894 - > Yes.

895 - > Most overhyped AI tool right now. 896 - > Uh open clock. 897 - > Okay, what's more dangerous? 898 - > Bad passwords or bad AI usage?

899 - > Bad AI usage. 900 - > One task every business should automate today. 901 - > Scheduling. 902 - > If a business owner only fixes one cybersecurity thing this 903 - > year, what should it be?

904 - > SPEAKER_02: Yeah. 905 - > SPEAKER_01: There's so many things they need to do. 906 - > There you go. 907 - > SPEAKER_00: Uh maybe awareness training.

908 - > That's a good one. 909 - > Yeah. 910 - > One AI tool you'd ban in a company. 911 - > Ooh.

912 - > SPEAKER_01: Um anything that gives system level access. 913 - > So, like claude co-work, for example. 914 - > SPEAKER_00: Claude co-work. 915 - > Best advice for someone scared of AI.

916 - > Just play with it. 917 - > Um try it in a browser. 918 - > That's more than one word, Brian. 919 - > Uh AI assistant or human assistant, which wins today?

920 - > AI assistant. 921 - > Wow. 922 - > Okay. 923 - > One word to describe AI right now.

924 - > One word. 925 - > Oh man. 926 - > We stumped him, ladies and gentlemen. 927 - > Evolving.

928 - > That's a good one. 929 - > And we'll wrap it up with one last one. 930 - > If AI were an employee, what would its job title be? 931 - > SPEAKER_01: Um intern.

932 - > SPEAKER_00: All right. 933 - > Do you have any AI jokes? 934 - > No. 935 - > SPEAKER_01: I could ask AI for an AI joke.

936 - > SPEAKER_00: All right. 937 - > Hey, look, man, you flew up here from Lakeland. 938 - > I appreciate it. 939 - > We're obviously we've got a lot of meetings this week, so 940 - > appreciate you coming up.

941 - > It's been too long. 942 - > Uh we see each other on Teams all the time, but uh we don't 943 - > get to hang out much in person. 944 - > So appreciate you coming up. 945 - > Thanks for being on Nerds on Tap again.

946 - > I honestly think this might be the best show we've ever done. 947 - > So I thank you. 948 - > Well, and you weren't tell the audience you were not prepped 949 - > for this show. 950 - > Not at all.

951 - > All right. 952 - > So are you sure? 953 - > You're not using AI right now to answer all these questions that 954 - > I just threw at you. 955 - > Nope.

956 - > What's that in your ear there? 957 - > My voice. 958 - > I'm just kidding. 959 - > Ladies and gentlemen, thank you for listening or watching 960 - > another episode of Nerds on Tap, where we like to get nerdy for 961 - > an hour.

962 - > Come back again and check out our next episode. 963 - > Thank you and make it a nerdy day. 964 - > Cheers, my fellow nerds and beer lovers. 965 - > Stay tuned for more Nerds on Tap.

966 - > Oh, and one more thing. 967 - > Help us spread the nerdy love and the love for grape brews by 968 - > sharing this podcast with your friends, colleagues, and fellow 969 - > beer enthusiasts. 970 - > Let's build a community that embraces curiosity, innovation, 971 - > and the enjoyment of a cold one.